Corinna M Bauer1, Lauren E Zajac2, Bang-Bon Koo3, Ronald J Killiany4, Lotfi B Merabet5. 1. Laboratory for Visual Neuroplasticity, Department of Ophthalmology, Massachusetts Eye and Ear Infirmary, Harvard Medical School, Boston, MA, USA. Electronic address: Corinna_Bauer@meei.harvard.edu. 2. Center for Biomedical Imaging, Department of Anatomy and Neurobiology, Boston University School of Medicine, Boston, MA, USA. Electronic address: zajac@bu.edu. 3. Center for Biomedical Imaging, Department of Anatomy and Neurobiology, Boston University School of Medicine, Boston, MA, USA. Electronic address: bbkoo@bu.edu. 4. Center for Biomedical Imaging, Department of Anatomy and Neurobiology, Boston University School of Medicine, Boston, MA, USA. Electronic address: killiany@bu.edu. 5. Laboratory for Visual Neuroplasticity, Department of Ophthalmology, Massachusetts Eye and Ear Infirmary, Harvard Medical School, Boston, MA, USA. Electronic address: Lotfi_Merabet@meei.harvard.edu.
Abstract
BACKGROUND: Deterministic diffusion tractography obtained from high angular resolution diffusion imaging (HARDI) requires user-defined quantitative anisotropy (QA) thresholds. Most studies employ a common threshold across all subjects even though there is a strong degree of individual variation within groups. We sought to explore whether it would be beneficial to use individual thresholds in order to accommodate individual variance. To do this, we conducted two independent experiments. METHOD: First, tractography of the arcuate fasciculus and network connectivity measures were examined in a sample of 14 healthy participants. Second, we assessed the effects of QA threshold on group differences in network connectivity measures between healthy young (n=19) and old (n=14) individuals. RESULTS: The results of both experiments were significantly influenced by QA threshold. Common thresholds set too high failed to produce sufficient reconstructions in most subjects, thus decreasing the likelihood of detecting meaningful group differences. On the other hand, common thresholds set too low resulted in spurious reconstructions, providing deleterious results. COMPARISON WITH EXISTING METHODS: Subject specific thresholds acquired using our QA threshold selection method (QATS) appeared to provide the most meaningful networks while ensuring that data from all subjects contributed to the analyses. CONCLUSIONS: Together, these results support the use of a subject-specific threshold to ensure that data from all subjects are included in the analyses being conducted.
BACKGROUND: Deterministic diffusion tractography obtained from high angular resolution diffusion imaging (HARDI) requires user-defined quantitative anisotropy (QA) thresholds. Most studies employ a common threshold across all subjects even though there is a strong degree of individual variation within groups. We sought to explore whether it would be beneficial to use individual thresholds in order to accommodate individual variance. To do this, we conducted two independent experiments. METHOD: First, tractography of the arcuate fasciculus and network connectivity measures were examined in a sample of 14 healthy participants. Second, we assessed the effects of QA threshold on group differences in network connectivity measures between healthy young (n=19) and old (n=14) individuals. RESULTS: The results of both experiments were significantly influenced by QA threshold. Common thresholds set too high failed to produce sufficient reconstructions in most subjects, thus decreasing the likelihood of detecting meaningful group differences. On the other hand, common thresholds set too low resulted in spurious reconstructions, providing deleterious results. COMPARISON WITH EXISTING METHODS: Subject specific thresholds acquired using our QA threshold selection method (QATS) appeared to provide the most meaningful networks while ensuring that data from all subjects contributed to the analyses. CONCLUSIONS: Together, these results support the use of a subject-specific threshold to ensure that data from all subjects are included in the analyses being conducted.
Authors: Bruce Fischl; David H Salat; Evelina Busa; Marilyn Albert; Megan Dieterich; Christian Haselgrove; Andre van der Kouwe; Ron Killiany; David Kennedy; Shuna Klaveness; Albert Montillo; Nikos Makris; Bruce Rosen; Anders M Dale Journal: Neuron Date: 2002-01-31 Impact factor: 17.173
Authors: Emily L Dennis; Neda Jahanshad; Jeffrey D Rudie; Jesse A Brown; Kori Johnson; Katie L McMahon; Greig I de Zubicaray; Grant Montgomery; Nicholas G Martin; Margaret J Wright; Susan Y Bookheimer; Mirella Dapretto; Arthur W Toga; Paul M Thompson Journal: Brain Connect Date: 2011
Authors: Nils Muhlert; Varun Sethi; Torben Schneider; Pankaj Daga; Lisa Cipolotti; Hamied A Haroon; Geoff J M Parker; Sebastian Ourselin; Claudia A M Wheeler-Kingshott; David H Miller; Maria A Ron; Declan T Chard Journal: J Magn Reson Imaging Date: 2012-12-12 Impact factor: 4.813